Download as pdf or txt
Download as pdf or txt
You are on page 1of 14

CLUES TO CREATIVITY

MAURITA HARNEY
Swinburne University a/Technology

To some, the topic of AI and Creativity (like the topic of 'machine thinking')
might sound like a contradiction in terms. For creativity in thinking and other activ-
ities, as commonly understood, requires us to withdraw the constraints implicit in
rules and rational appraisal. But what could be more rational or rule-bound than the
model of the mind and cognition on which AI is founded? The implicit connection
here, between creativity and 'irrationalism', owes something to Sigmund Freud,
who saw creativity as the link between art and play. Both are activities pursued
for their own sake, and both involve the suspension of rational principles: "The
creative writer does much the same as the child at play. He creates a world of phan-
tasy which he takes very seriously-that is, which he invests with large amounts of
emotion-while separating it sharply from reality.',1 There is no reason to suppose
that creativity in science is any different. It is identified as that initial phase in
scientific inquiry when the principles of rational evaluation and assessment are sus-
pended and a bold conjecture or intuitive guess is made. It is that phase prior to the
process of testing by means of the rational principles of deduction and induction.
When creativity is identified with thinking which is intuitive, conjectural or
hypothetical, rather than rule-governed or rational, it is tempting to dismiss it as
'irrational'-as something mysterious, ineffable, resistant to attempts to analyse it
or appraise it in any way. However, I prefer to dispense with the dubious dichotomy
of rationaVirrational, and focus instead on the kinds of inquiry which might be called
conjectural, creative, or intuitive. I will be suggesting that conjectural knowledge of
this kind constitutes an epistemological paradigm (in Kuhn's (1970) sense 2 ) which
is different from, but complementary to, the paradigm of scientific knowledge
(which is governed by considerations relating to rational explanation, and notions
of evidence, truth, and justification). It is a paradigm more easily recognisable in
the study of cultural meanings than in standard approaches to the study of scientific
phenomena. For this reason I will be suggesting that we tum to the domain of
cultural inquiry to seek the conceptual tools for analysing the creative component
in scientific discovery. More specifically, I propose to explore some ideas drawn
from recent developments in literary theory and studies in rhetoric. It is here that I
think we might find some valuable insights for conceptual ising the creative process.
Such an exploration is itself an exercise in creativity, for it requires us to transgress
1 Freud, S.: 1959, Creative Writers and Day-Dreaming, standard edn, Vol. IX, Hogarth Press and
the Instirute of Psychoanalysis, London, p. 144, quoted in Storr (1972: 113). Margaret Boden (1990:
46) similatly points out that "creativity has much in common with play".
2 The notion of a conjectural paradigm, however, is derived from Carlo Ginzburg in Eco and
Sebeok (1983).

195
T. Dartnall (ed.). Artificial Intelligence and Creativity, 195-208.
© 1994 Kluwer Academic Publishers.
196 MAURITA HARNEY

the traditional boundaries which demarcate and dichotomise the arts versus the
sciences, the factual versus the mythical, the rational versus the poetic.

The American philosopher C. S. Peirce (1901 ff.) provides us with a useful starting
point for talking about knowledge which is conjectural, creative, or intuitive. He
coined the tenn 'abduction' (variously called 'retroduction' and 'hypothesis') to
describe that initial phase in scientific inquiry when a bold hypothesis is fonned to
explain some surprising fact or observation.
Peirce regarded Kepler's discovery of the elliptical orbit of Mars to be a classic
case of abduction. Kepler's reasoning was as follows: First, the surprising fact of
certain irregularities in the path of Mars is observed. But, if the orbit of Mars were
elliptical, then these movements would be a matter of course. So, hypothetically,
Mars travels in an elliptical orbit:

A surprising fact, C, is observed.


But, if A were true, then C would be a matter of course.
So, hypothetically,A.

Or, schematically:

C
If A then C
So, A

Of course, this hypothesis can only be entertained tentatively, awaiting testing by


further observation and reasoning. This phase in scientific inquiry is where the
creative element enters. It has often been described as conjecture-the result of
intuition, a leap of the imagination, or a guess. It contrasts with the processes of
reasoning (deduction and induction) whereby we test orjustify hypotheses.
Elsewhere, Peirce (1878) spells out the distinctive features of abduction. This
occurs in his earlier writings where he characterises abduction (or 'hypothesis' as
he calls it) as a mode of reasoning alongside, but different from, deduction and
induction. 3
In a deductive argument, the premises imply the conclusion, and the inference
is non-ampliative. It is the application of a rule to a particular case:

If A then C
A
So,C
3 See Tursman (1987) for a detailed discussion of Peirce's logic.
CLUES TO CREATIVITY 197

In contrast, abduction is not the application of a rule to a case. If it were it would


be a deductive fallacy-the fallacy of affirming the consequent. Unlike the case of
valid deduction, true premises in an abductive inference do not guarantee the truth
of the conclusion. The conclusion of an abductive inference can only be held as a
hypothesis awaiting further testing. As Peirce says, it must be held interrogatively,
as a question. We cannot affirm its conclusion with certitude or confidence. It
involves an element of risk.
However, correlated with this element of risk is an element of inventiveness or
originality. Abduction introduces something different. It creates new knowledge.
This distinguishes it from induction, which is a generalisation, or a continuation of
a series.
To illustrate this difference, Peirce (1878) gives us the following example. We
have a torn piece of paper with some anonymous writing on it, and we suspect a
certain person to be the author. When we search that person's desk, knowing that
only he has access to it, and find another torn piece of paper which matches exactly
the one we hold, it is a reasonable hypothesis to suppose that the suspected man
was in fact the author. The inference here, argues Peirce, is not an induction but an
abduction. Induction would only warrant our inference that the two pieces of paper
which matched in certain respects so far examined, would exhibit further matchings
of the same kind. "The inference from the shape of the paper to its ownership is
precisely what distinguishes hypothesis (i.e. abduction) from induction and makes
it a bolder and more perilous step." (p. 140).
In summary, then, the main features of abduction which emerge from Peirce's
earlier writings about abduction as a logical category are the following:
Abduction must always be entertained as a question. That is, it must await
further testing by deduction or induction before we can hold it with confidence.
Abduction involves an element of risk.
Abduction is an inference to something different, rather than to a generalisa-
tion.
Abduction involves an element of inventiveness or creativity.
Abduction is an inference from an effect to a cause, not (as induction is) merely
reasoning from particulars to the general law: induction merely classifies;
abduction explains.
Peirce offers us a model of the cognitive process invol ved in scientific discovery,
but his account does little by way of suggesting answers to the questions: How is
the new hypothesis formed? What constitutes a 'surprising fact'?
Computational applications of Peirce's notion don't seem to help us greatly here,
either. Abduction is currently used in the design of expert systems, particularly in the
area of diagnostic hypothesis generation and in explanation-based learning systems
(O'Rorke, 1990; Punch, Tanner, Josephson and Smith, 1990) where it is the basis
of a kind of backward chaining from symptoms or effects to a plausible hypothesis
about the causes of those phenomena. In this context, abduction is closely linked
with inference to the best explanation (Harman, 1965; Achinstein, 1971). Because
198 MAURITA HARNEY

the process of discovery is not separated from the process of hypothesis evaluation,
the sense in which abduction might be called 'creative' is, to say the least, both
philosophically and computationally uninteresting. Paul Thagard (1988) has used
abduction as a computational mechanism in developing a model of the process of
knowledge-acquisition. He describes a program which uses abduction (along with
analogy and generalisation) to generate the concept 'sound-wave'. This has not been
greeted with enormous enthusiasm-possibly because of the very restricted sense
of creativity that it involves. It works by matching preselected concepts so that its
'creativity' ranking, in the eyes of one reviewer (Schagrin, 1991), is equivalent to
that of completing a multiple-choice test. Moreover, as Thagard himself admits, his
program doesn't take into sufficient account the background knowledge to scientific
conjecture.

Peirce himself did not develop a single unified or consistent theory of abduction.
Nevertheless he presented us with an enormously rich and fertile notion which
others have subsequently developed more fully in a vast range of cognitive domains
dealing with human inquiry. It has, for example, been identified with the mode
of reasoning used by fictional detectives like Sherlock Holmes (notwithstanding
Conan Doyle's claim that the method was one of observation and deduction).
It is a notion which has figured prominently in the writings-both fictional and
philosophical-of Umberto Eco. Eco has an impressive record of scholarship in
the philosophy of language and rhetoric. He is, of course, popularly known for his
fictional works and screen writing as well.
Anyone who is familiar with Umberto Eco's mediaeval detective story, The
Name of the Rose, may have recognised in its opening pages a variation on ano):her
piece of fiction-Voltaire's fairy tale of Zadig. This story tells how Zadig, a very
astute observer of 'the book of nature', was walking in the forest one day when
he met two bands of very agitated men. They were the queen's courtiers and they
were anxiously searching for the queen's horse and her pet dog which had recently
escaped. They asked Zadig if he had seen them. Zadig asked if the horse was a
stallion with a perfect gallop and small hoofs, so many hands high, with a three and
a half foot tail, a gold studded bit and silver shoes. He went on to ask if the dog
was a spaniel bitch which had long ears and an injured left foreleg, and which had
recently given birth to pups. His descriptions were so accurate that the courtiers
exclaimed that he must have seen the horse and the dog. When Zadig denied this,
and said that he had never set eyes on them, the courtiers understandably refused
to believe him. Not only that, they presumed that he must have stolen the beasts,
and promptly dragged him off to prison.
Later, Zadig was able to tell his story. Being a sharp observer of his natural
surroundings, he had perceived several signs-imprints in the soil, a broken branch
and dust on the leaves at a certain height, a trace of gold on a rock, furrows in
the dust, and so forth-and from these had built up a general picture of what had
CLUES 10 CREATIVITY 199

passed by. He had, in other words, formed a hypothesis about what had occurred
on the basis of keenly observed signs.
This same kind of reasoning from observed signs to a general picture of what
caused those signs is the method of investigation used by fictional crime detectives
like Sherlock Holmes, and indeed by Umberto Eco's own fictional sleuth in his
novel, The Name a/the Rose. Here again, in detective fiction, we have the careful
observation of signs or 'clues'-a footprint, a forced lock, a blood-smeared poker,
etc.-from which the detective formulates a hypothesis or general picture of the
possible origins of those clues. In crime detection, this general picture is very
complex, and includes things like motives, intentions, etc.

Abduction is a particularly useful mechanism in explaining knowledge acquisi-


tion in areas where empirical methods for testing hypotheses are not available-
hypotheses about past events or events which are unique, for example. Peirce
himself gives the example of our historical knowledge of the fact that Napoleon
once lived and performed various feats. According to Peirce, this knowledge is
based on abduction. Numerous documents and monuments refer to a conqueror
named 'Napoleon'. We have never encountered the man. However, to explain these
observed monuments and documents, we must hypothesise the reality of his exis-
tence. For Peirce, knowledge of the physical world is derived in exactly the same
way: for example, fossil remains of fish are found far in the interior of a continent.
To explain this phenomenon, we hypothesise that the sea once covered this land
(peirce, 1878: 135, 150-151).
Abductive processes are also evident in our methods for gaining knowledge
about the authorship of works of art. The cultural historian Carlo Ginzburg (1983)
shows how the work of art authentication undertaken by Giovanni Morelli was based
on this method. Morelli's method in authenticating works of art or exposing imita-
tions consisted in using seemingly insignificant details of a painting-fingernails,
earlobes, etc.-rather than the main subject or the obvious stylistic conventions-as
clues from which to draw up hypotheses about the authenticity of works of art. (This
method came to be known as 'Morelli's method' in art history circles). Ginzburg
goes on to identify similarities between the methods of Morelli, Freud and Sherlock
Holmes, all of whom use seemingly trivial details as clues from which a larger,
more complex picture is constructed. He calls this the 'conjectural model for the
construction of knowledge' ,4 and suggests that the origins of this method of inquiry
are traceable back to the very earliest methods of knowledge-acquisition used by
hunters and trackers, who 'read' animal droppings, footprints, feathers, hairs, etc.,
as signs of past events. It is a method of both diagnosis and prognosis. He points out
4 He suggests that conjectural knowledge constitutes an epistemological paradigm (in Kuhn's
sense) which "quietly emerged towards the end of the nineteenth centul)' in the sphere of the social
sciences, and still has not received the attention it deserves (p. 81)
H

200 MAURITA HARNEY

evidence of this method in the work of Mesopotamian diviners, for whom animals'
innards, stars, drops of oil in water, etc., were read as signs of the future. Ginzburg
(1983: 110) remarks:
It's a matter of kinds of knowledge which tend to be unspoken, whose rules
... do not easily lend themselves to being formally articulated or even spoken
aloud. Nobody learns to be a connoisseur or a diagnostician simply by applying
the rules. With this kind of knowledge there are factors in play which cannot
be measured-a whiff, a glance, an intuition ... 5
The suggestion is that conjectural knowledge constitutes an epistemological
paradigm different from that of scientific knowledge. Scientific knowledge is gov-
erned by considerations of justification, testing, evidence, and notions of truth and
proof. The conjectural paradigm gives primacy to processes of discovery rather
than justification or evaluation, and consists of knowledge that is often regarded as
pre-scientific or even pseudo-scientific. Within this paradigm, the discovery process
is cast as a theory about interpreting signs. Abduction is the underlying method of
interpretation and for this reason constitutes, in Eco's words, a semiotic mechanism.

Semiotics means the science of interpreting signs. Like cognitive science, it has
emerged as an interdisciplinary field of study in recent years. Its focus of inquiry
is cultural meanings. All cultural phenomena (institutions, mythS, dress, sport,
film, advertising, media, etc.) are seen as signs or sign systems whereby meaning
is communicated. Semiotics draws on linguistics, philosophy, literary theory, and
studies in rhetoric for the methods and techniques for 'decoding' or 'deciphering'
these signs.
As a way of re-conceptualising the process of forming bold hypotheses, I will
be suggesting that the abductive phase in scientific inquiry be re-located within
the alternative epistemological paradigm described by Ginzburg as the 'conjectural
paradigm'. Within this context, abduction can be theorised as a semiotic mechanism,
that is, a device used in the interpretation of signs. This might, incidentally, be
regarded as an extension of Peirce's own insights, as Peirce himself is one of the
acknowledged founders of semiotics.
Quoting Peirce's own words, Eco (1984) describes a sign as "something by
knowing which we know something else".6 Eco's notion of 'sign' is articulated
through a careful historical analysis of the notion beginning with Hippocrates'
idea of a sign as a symptom. It is this notion of sign which is operative when
we recognise smoke as a sign of fire. This recognition is abductive in structure.
Smoke only becomes a sign when the interpreter sees the phenomenon/event as
5 Ginzburg's examples of 'conjectural' knowledge are reminiscent of Polanyi's (1962) 'tacit
knowledge'. The conceptual frameworks in each case are different, however.
6 The reference is to Peirce's Collected Papers, 8.332.
CLUES TO CREATIVITY 201
the antecedent of a hypothetical inference of the fonn ifp then q. Understanding
linguistic meaning has precisely the same structure. Our understanding of the
utterance /man/ as meaning rational, mortal animal is not modelled on equivalence
or identity. (It is not of the fonn p if and only if q).7 It is not a function of
substitutability of synonymous tenns. Rather, it is based on hypothesis in the same
way that' Smoke means fire' is. My understanding assumes first that the utterance
/man/ is an utterance (token) of a type of English word: "To recognise a given
phenomenon as the token of a given type presupposes some hypothesis about the
context of utterance and the discursive co-text" (Eco, 1983: 206).
Eco's abductive analyses of such commonplace phenomena are quite consistent
with Peirce's own views. Peirce believed that abduction pervades all of our mental
life, from the most commonplace identifications through to truly revolutionary
scientific discoveries. Perception is a case in point. If certain perceptual data are
present then there is perhaps an inkwell, "as long as other contextual elements
authorise me to think that the perceptual interpretation is appropriate" (Eco, 1984:
35). Taking up Peirce's notion that perception is 'presumptive evidence', Eco (1984)
says: "Perception is always interrogative and conditional and is invariably based
(even if we do not realise it) on a bet ... Perception is ... a source of potential semiosis.
The fact that perception takes place without effort does not invalidate its inferential
mechanism."8 Such inferences are, of course, so commonplace as to be quasi-
automatic. They are what Eco calls 'overcoded abductions', in which the rule (the
'if... then... ' clause) is 'already coded'. This kind of abduction lies at one end of a
spectrum. At the other end are what Eco calls the 'creative abductions' exemplified
in revolutionary scientific discovery like Einstein's Theory of Relativity. Here, the
abduction involves a rule which is invented ex novo. In between are abductions
exemplified by what might be called 'creative problem solving'. Here the rule
is selected from a number of equiprobable rules at our disposal. They are called
'undercoded abductions', and include the kind of inquiry undertaken by fictional
detectives when they reason from observed clues to a hypothesis about 'whodunit',
or the reasoning in routine medical diagnosis.
The further we move along this spectrum, the more creative the hypothesis; the
more creative the hypothesis, the riskier it is.

The process of reasoning which leads to a conjecture or hypothesis has been likened
to the construction of a narrative or plot. A plot, unlike a chronicle or mere list of
events, is a coherent, schematic whole (polkinghome, 1988: 19):9
T Bco's insistence on reserving the hypothetical (if. .. then ... ) relation as the model for sign·
inteIpretation, and distancing it from the equivalence (... if and only if. .. ) relation (represented by
synonymy) preserves Peirce's important logical distinction between abduction and deduction.
8 Bco's reference is Peirce's CoUected Papers, 5.266-68.
9 Polkinghome's work includes a study of ways in which narrative is treated as a cognitive structure.
See 1988, chapter 5.
202 MAURITA HARNEY

A plot is able to weave together a complex of events to make a single story.


It is able to take into account the historical and social context in which the events
took place and to recognize the significance of unique and novel occurrences
... The reasoning used to construct a plot is similar to that used to develop a
hypothesis. Both are interactive activities that take place between a conception
that might explain or show a connection among the events and the resistance of
the events to fit the construction.
Polkinghorne, who draws this connection between hypothesis and plot construc-
tion, suggests that narrative understanding constitutes a unique mode of explanation,
different from that provided by 'logico-mathematical reasoning' which is normally
operative in scientific inquiry. What is distinctive about narrative understanding
is the dialectical nature of the relationship between event and plot. What makes
something a plot structure rather than a mere chronicle of events is the significance
of those events and the relationships between them. But the meaning or identity of
a singular event is itself a function of the interaction between the event and the plot
structure. Plot construction is
not the imposition of a ready-made plot structure on an independent set
of events; instead it is a dialectic process that takes place between the events
themselves and a theme which discloses their significance and allows them to
be grasped together as parts of one story. In addition the construction of plots
is not a completely rule-governed activity. It can generate unique and novel
configurations (polkinghorne, 1988: 19-20).
Within the context of scientific reasoning, it is common to suppose that inventing
the bold hypothesis, A, on the basis of an observed fact, C, is a matter of identifying
some kind of general law ('Whenever A then C') which can be fitted over the
individual facts or events (A and C), on the assumption that the latter can be treated
as independent, context-free elements. "The power of explanation by laws comes
from its capacity to abstract events from particular contexts and discover relation-
ships that hold among all the instances belonging to a category, irrespective of the
spatial and temporal context" (polkinghorne, 1988: 21). However, 'subsumption
under a general law' is far too restrictive a way of characterising the relationship
between C and the hypothetical 'if A then C', if we are to do justice to the creative
element in hypothesis-construction. (Although it is essential for the subsequent
task of testing those hypotheses.) And, indeed, it is contrary to the spirit of Peirce's
own account, for he regarded subsumption under a rule as being characteristic of
deduction rather than abduction. Semiotics, however, provides us with a way of
preserving Polkinghorne's insights concerning the distinctive features of narrative
understanding and its relevance to creativity in science. 10 Semiotics provides us
with a theoretical framework which might be regarded as an alternative to that of the

10 Polkinghome's own approach is not semiotic. However, his approach is consistent with the more
general project of drawing on the insights ofliterary theory to make sense of the nature of knowledge
and inquiry in other, seemingly unrelated, domains.
CLUES TO CREATNITY 203

law-governed explanatory schema commonly associated with scientific reasoning.


It is a framework that is constituted by a rich array of conceptual and methodolog-
ical tools drawn from literary theory. A semiotic relationship between C and A can
be conceptualised in a range of possible ways, none of which is reducible to the
notion of a generalisation in the sense outlined above. 1 1 C can be a symptom of A,
a clue to A, a token of some type, A, an attribute of A. Recognising the relationships
in question depends on analogy with experience. But here, again, the conceptual
framework of scientific reasoning is too restrictive to do justice to the rich array of
analogical and associative thinking that is possible.
The process of coming up with a bold hypothesis depends on analogy with past
experience. A distinction is sometimes made between 'argumentative analogy' and
'non-argumentative analogy'. The former is a kind of induction which consists in
recognising properties which are common to two sets of phenomena (e.g. that both
sound and waves propagate and reflect) and then drawing the inference that those
two things are common in other respects. This account of analogy is far too restricted
to provide an adequate basis for conceptualising the creative process. Moreover, it
would reduce abduction to induction, thereby negating Peirce's original claim. Non-
argumentative analogy is exemplified by associative thinking rather than thinking
based on inference. A good example is Freud's technique of free association in
which the analyst utters a word and the client responds with the first thing that
comes to mind. The result is a series of ideas or images linked not by logic but by
association of ideas. This then becomes the basis for building up a coherent general
picture or explanation of the client's symptoms.
This notion of non-argumentative analogy derives from metaphor, which is the
association of two things which are not conventionally linked, for example when
we call New York 'The Big Apple' or when we say of a ship that it 'ploughed the
deep'.
Semiotics provides us with a vast repertoire of devices fortheorising the different
kinds of associations (or non-argumentative analogies) exhibited in the interpreta-
tion of signs. Metaphor is just one of them. Others include metonymy and synec-
doche, which are drawn from literary theory and studies in rhetoric. A brief look at
how these devices work will help us to see how the notion of non-argumentative
analogy might be theorised. 12
Semiotics makes use of a structural framework drawn from linguistic theory
to show how linguistic messages are constructed and interpreted. Its starting point
is the idea that "speech implies a selection of certain linguistic entities and their
combination into linguistic units of a higher degree of complexity" (Jakobson,
1956, cited in Lodge, 1977: 74). These two operations of selection and combination
are represented by two axes: one, a vertical or paradigmatic axis representing the

11 Schank (1990) uncritically conflates the ability "to connect together experiences which are not
obviously connectable" with "generalisation", claiming that the latter is essential to creativity.
12 There are differences amongst semioticians as to the application of these distinctions. These
differences need not concern us here.
204 MAURITA HARNEY

selection of a linguistic unit; the other a horizontal or sYlltagmatic axis, representing


the combination of the selected units.

selective (paradigmatic) dUDensior.


metaphor

1
~ _,~,,~ ,.~._"., .~,,~"
metonymy

The selection process is governed by relations of similarity; the combination


process by relations of COllfigllity. Lodge (1977) gives the example of the sentence
•Ships crossed the sea'. This message is constructed by (a) the selection of certain
linguistic units from a set of similar kinds of things ('ships' instead of 'craft',
'vessels" etc.; 'crossed' instead of 'traversed', 'went over', etc ...), and (b) the
combining of these units by placing them next to each other in a certain order.

SYNTAGM ~ Ships crossed the sea

1
boats traversed the water

craft went over the ocean

vessels sailed etc.

PARADIGM etc. etc.

Anything which communicates meaning is constructed by means of these two


processes of selection and combination. A single word is a syntagm constructed
first by selecting each letter from a set of similar items (the alphabet), and then by
the process of combining those letters into a sequence according to the particular
linguistic conventions. The opposition between these two axes is fundamental to a
semiotic analysis of all cultural phenomena (Fiske, 1982: 62):
Our clothes are a syntagm of choices from the paradigms of hats, ties,
shirts, jackets, trousers, socks, etc. The way we furnish a room is a syntagm of
choices from the paradigms of chairs, tables, settees, carpets, wallpapers, etc.
An architect designing a house makes a syntagm of the styles of door, window,
and their positions. A menu is a good example of a complete system. The
choices for each course (the paradigms) are given in full: each diner combines
them into a meal; the order he gives the waiter is a syntagm.
The selection process represented by the vertical paradigmatic axis is the process
by which metaphor is generated. It implies the possibility of substitution based on
a certain kind of perceived similarity. In the sentence 'Ships ploughed the sea',
the word 'ploughed' is a metaphor for 'crossed'. The substitution of 'ploughed'
for 'crossed' is governed by the principle represented in the vertical axis. What
CLUES TO CREATIVITY 205
sanctions the substitution here is the perceived similarity between the movement of
the ship and that of a plough. The movement of the ship and that of the ploughshare
belong to the same 'set' ofthings, and are thus available alternatives for selection.
The combining of linguistic units, as represented by the horizontal syntagmatic
axis, is the process by which metonymy is generated. Metonymy is the substitution
of a part for the whole, of cause for effect, of attribute for the thing itself, of genus for
species and vice versa, of the concrete for the abstract. Metonymy is used in 'the big
smoke' to refer to a large industrial city, in 'the deep' to mean 'the sea'. We speak
of factory employees as 'hands', and the monarchy as 'the crown'. The principle
which governs metonymic substitutions derives from the relation of contiguity or
adjacency between, for example, an object and its attribute, or a part and the whole.
'Hands' can stand for factory workers not because of any similarity between the
two things (as in the substitution of 'ploughed' for 'crossed'), but because of the
relation of contiguity between them. Metonymy is based on space-time contiguity
and thus generates meaning according to the principle governing the horizontal
axis.
The metaphor/metonymy polarities are used as classificatory devices in a range
of cultural phenomena. In film, for example, the technique of montage is metaphoric,
while the close-up is metonymic; cubist painting is metonymic, whilst surrealism
is metaphoric. In Freudian interpretation of dreams, the processes of condensation
and displacement are metonymic, while the processes of symbol-formation and
identification are metaphoric.
Some literary theorists make a further distinction, using synecdoche for the
'part/whole' substitution, and reserving metonymy for the process of substituting
an attribute for the thing itself. Thus, 'keel' is a synecdoche for 'ship', and 'the
deep' is a metonym for 'the sea'. These distinctions are not mutually exclusive: the
metaphor/metonymy polarity can occur within the metaphorical mode or within
the metonymic mode. The sentence 'Keels ploughed the deep' is a metaphor for
ships crossing the sea. But within this metaphor, synecdoche is used in substituting
'keel' for 'ships' (part for whole), 'ploughed' is a metaphor for 'crossing', and
metonymy is used in the substitution of 'deep' for 'the sea' (attribute for the thing
meant) (Lodge, 1977).
Equipped with this array of devices for classifying different kinds of associa-
tions, we can perhaps re-conceptualise the processes of abduction described earlier.
The inference made on the basis of imprints might be seen as involving a complex of
associative relations-for example, a metonymic relationship between imprint and
hoof (effect/cause), synecdochical relations between hoof and horse (part/whole)
and between horse and stallion (genus/species). The token-type relationship un-
derlying perceptual recognition might be seen as synecdoche. Anything which is
identified as a clue or a symptom (and hence as a sign) might be similarly classified
according to the precise nature of its relationship to its origins.
It would be a mistake to dismiss these notions as 'mere literary devices'. Studies
in aphasia (a speech disorder) undertaken by linguist Roman Jakobson suggest that
206 MAURITA HARNEY

metaphor and metonymy represent fundamental cognitive structures or orientations


(Jakobson and Halle, 1956). 13 Jakobson's approach marks a departure from tradi-
tional approaches, which studied aphasia in tenns of a distinction between sending
and receiving verbal messages. By contrast, Jakobson "makes his methodological
'cut' along the line between selection and combination" (Lodge, 1977: 77). He
distinguishes two kinds of aphasia, one resulting from a deficiency in the syntag-
matic function (contiguity disorder), the other resulting from a deficiency in the
paradigmatic function (selection disorder). His studies show how subjects suffering
from speech disorders in the syntagmatic mode tend to compensate by transferring
the paradigmatic operation to the syntagmatic axis. The reverse happens in subjects
suffering from disorders in the paradigmatic mode. So, for example, patients suf-
fering from the contiguity disorder would typically have difficulty with word-order
and syntax. Grammatical connectives would disappear, and sentences would be re-
placed by single-subject words. Moreover, as if to compensate for this metonymic
deficiency, the patient would use metaphorical substitutions such as 'spyglass' for
'microscope', or 'fire' for 'gaslight'. On the other hand, an aphasic patient suf-
fering from a deficiency in the paradigmatic function (what Jakobson calls the
'selection disorder'), exhibits difficulties in dealing with relationships governed
by the selection principle-names, definitions, synonymies. Although word-order
is preserved, the patient's speech appears to be confined to context-dependent ex-
pressions. In these patients, the metonymic function dominates discourse and is
even projected onto the selection axis. So, for example, objects are identified by
metonymic substitutions-'fork' for 'knife', 'table' for 'lamp', 'smoke' for 'pipe'.
That is to say, something (e.g. 'table') which usually occurs alongside, or is con-
tiguous with, something else is selected in place of that thing.
Lodge (1977: 78-79) adds the observation that "if much modem literature is
exceptionally difficult to understand, this can only be because of some dislocation
or distortion of either the selection or combination axes of language; and of some
modem writing, e.g. the work of Gertrude Stein and Samuel Beckett, it is not an
exaggeration to say that it aspires to the condition of aphasia".
In conclusion, then, I am suggesting that there is a way of re-conceptualising
the creative or abductive phase in the process of scientific inquiry. We can re-
locate the notion of abduction within the conjectural paradigm and theorise it as a
semiotic mechanism. This furnishes us with an impressive range of devices drawn
from literary theory and studies in rhetoric for classifying the various kinds of
associative relations which enable a hypothesis to be fonned on the basis of a
surprising fact. We have, in short, a greatly enriched set of conceptual tools for
analysing the 'if... then .. .' relationship which constitutes the bold hypothesis, and
for representing the mechanisms by which such hypotheses are generated.

13 Jakobson's 'structuralist' approach to the relationship between cognition, linguistics and rhetoric
has established his reputation as a forefather of modem semiotics. 'Non-structuralist' approaches to
the study of metaphor and its relation to cognition include the work of Lakoff (1987), and Lakoff and
Johnson (1980).
CLUES TO CREATIVITY 207
References
Achinstein, P.: 1971, Law and Explanation: An Essay in the Philosophy o/Science, Clarendon Press,
Oxford.
Boden, M.: 1990, The Creative Mind: Myths and Mechanisms, Weidenfeld and Nicolson, London.
CUller, J.: 1981, The Pursuit o/Signs: Semiotics, Literature, Deconstruction, Routledge and Kegan
Paul, London.
Eco, U.: 1976, A Theory o/Semiotics, lndiana University Press, Bloomington.
Eco, U.: 1979, The Role o/the Reader, Indiana University Press, Bloomington.
Eco, U.: 1983, Horns, hooves, insteps, in, Eco, U. and Sebeok, T. A. (eds), The Sign o/Three: Dupin,
Holmes, Peirce, Indiana University Press, Bloomington.
Eco, U.: 1984, Semiotics and the Philosophy o/Language, lndiana University Press, Bloomington.
Eco, U. and Sebeok, T. A. (eds): 1983, The Sign of-Three: Dupin, Holmes, Peirce, lndiana University
Press, Bloomington.
Eco, U.: 1990, The Limits 0/ Interpretation, lndiana University Press, Bloomington.
Eco, U., Santambrogio, M. and Violi, P. (eds): 1988, Meaning and Mental Representation, University
of Chicago Press, Chicago.
Fiske,J.: 1982, Introduction to Communication Studies, Methuen, London.
Ginzburg, C.: 1983, Morelli, Freud and Sherlock Holmes, in Eco, U. and Sebeok, T. A. (eds), The
Sign o/Three: Dupin, Holmes, Pdrce, lndiana University Press, Bloomington, pp. 81-118.
Harman, G.: 1965,lnference to the best explanation, Philosophical Review, 64: 88-95.
Hawkes, T.: 1977, Structuralism and Semiotics, Methuen, London.
Hofstadter, D. R: 1985, Metamagical Themas: Questingfor the Essence 0/ Mind and Pattern, Basic
Books, New York.
Jakobson, R: 1956, Two aspects of language and two types of linguistic disturbances, in Jakobson,
R. and Halle, M. (eds), Fundamentals o/Languages, Mouton, The Hague.
Jakobson, R. and Halle, M. (eds): 1956, Fundamentals o/Language, Mouton, The Hague.
Johnson-Laird, P. N.: 1988, The Computer and the Mind, Fontana, London.
Kuhn, T.: 1970, The Structure 0/ Scientific Revolutions, University of Chicago Press, Chicago.
Lakoft', G.: 1987, Women, Fire, and Dangerous Things, University of Chicago Press, Chicago.
Lakoft', G. and Johnson, M.: 1980, Metaphors We Live By, University of Chicago Press, Chicago.
Lodge, D.: 1977, The Modes 0/ Madern Writing: Metaphor, Metonymy, and the Typology 0/ Madern
Literature, Edward Arnold, London.
O'Rorke, P. (ed.): 1990, Working Notes of the 1990 Spring Symposium on Automated Abduction,
Technical Report 90-32, University of California, Irvine.
Partridge, D. and Wilks, Y. (eds.): 1990, The Foundations 0/ Arrijicial Intelligence: A Sourcebook,
Cambridge University Press, Cambridge.
Peirce, C. S.: 1878, Deduction, induction, and hypothesis, in Cohen, M. R (ed.), 1923, Chance, Love,
and Logic, Kegan Paul, Trench, Trubner, pp. 131-153.
Peirce, C. S.: 1901, Abduction, induction and deduction, in Burks, A. W. (ed.), 1966, Collected
paperso/Charles Sanders Peirce, Vol. VII, BelKnap Press, Cambridge Mass.
Peirce, C. S.: 1901ft'., Abduction and induction, combining selections from mss of 1901, 1903, 1896
and 1908, in Buchler, J. (ed.), The Philosophy 0/ Peirce. Selected Writings, Routledge & Kegan
Paul, London, 1940, pp. 150-56.
Polanyi, M.: 1962, Personal Knowledge: Towards a Post-critical Philosophy, Routledge & Kegan
Paul, London, corrected version 1962.
Polkinghorne, D. E.: 1988, Narrative Knowing and the Human Sciences, State University of New
York Press, Albany.
Punch, W. F., Tanner, M. C., Josephson, J. R, and Smith, J. W.: 1990, Peirce: A toolfor experimenting
with abduction, IEEE Expert, October: 34-44.
208 MAURITA HARNEY

Schagrin, M. L.: 1991, Review of Paul Thagard, Computational Philosophy of Science, Minds and
Machines, 1(1): 121-124.
Schank, R.: 1990, What is AI anyway'! in Partridge, D. and Wilks, Y. (eds), The Foundations of
Artificial Intelligence: A Sourcebook, Cambridge University Press, Cambridge, pp. 3-13.
Storr, A.: 1972, The Dynamics of Creation, Seker and Warburg, London.
Thagard, P.: 1988, A Computational Philosophy of Science, MIT Press, Cambridge, MA.
Tursman, R.: 1987, Peirce's Theory of Scientific Discovery, Indiana University Press, Bloomington.

You might also like